We propose to develop a new series of methods for family studies in behavior disorders such as bipolar affective disorder (manic-depressive psychosis) and schizophrenia. These methods are specifically designed to detect the effect of single genes in the presence of four known complicating factors: reduced penetrance, assortative mating, association, and ascertainment bias. Current family study techniques do not take account of these complications, which are, however, known to occur in behavioral traits. The techniques we will develop represent a logical continuation of twin and adoption studies previously used to define the relative importance of the genetic and environmental contributions to the etiology of behavior disorders. We base our proposed methods on a likelihood analysis of extended families using only information on affected individuals. This is an extension of the concept of analyzing only affected siblings. Development of our methods involves defining for an analysis the extent of the problem posed by the four complications, devising appropriate techniques to circumvent each problem, and verifying the techniques through formal mathematics and an analysis of simulated data. We will then apply our new methods to a data set of approximately 150 families ascertained through manic-depressive probands in order to compare our results with those obtained by standard analysis.